ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
2

Fusing multiple sensors of same type using robot_localization.

asked 2018-03-14 04:01:25 -0500

Salahuddin_Khan gravatar image

Hi, I understand fairly clearly how data from Multiple sensors of different type like IMU, GPS , Wheel-Odometry etc can be fused using the Kalman-Filter.

But how does robot_localization fuses data from multiple sensors of same type say velocity input from 2 Wheel Odometers?

Is it simply weight averaged using their individual co-variances and the combined output is sent as a single Odometry measurement to the Kalman filter or is it a lot more complex process?

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted
2

answered 2018-03-14 06:58:27 -0500

tuandl gravatar image

According to line 175, if you are sure about your covariances report from both wheel odometry, then each velocity measurement will be fed to the KF individually based on its time stamp. Make sure that you follow the documentation before doing so.

edit flag offensive delete link more

Comments

I see, so at each iteration of Kalman Filter, Velocity measurement from just one Odometer is given as input based on the timestamp and a different H matrix is created depending on which sensor data is being fused. This Helps. Thanks!

Salahuddin_Khan gravatar image Salahuddin_Khan  ( 2018-03-15 03:45:50 -0500 )edit
2

answered 2018-03-14 12:36:59 -0500

Gayan Brahmanage gravatar image

Hi, In order to perform SLAM or Localization, you need both Proprioceptive sensors (Internal state like an odometer, IMU, wheel encoder) and Exteroceptive sensors (external state like laser scanners). You need to use internal measurements to perform prediction with a given motion model. For correction step, you need to use external measurements like laser scans.

If you use wheel encoder system, you need a mathematical model to predict the next state of the robot in the prediction step of the EKF. you can use two velocity inputs or counts from wheel encoder as control inputs for state transition equations (prediction step). For example, two-wheel differential-drive-mobile robot. Then you can do correction using external measurements.

edit flag offensive delete link more

Comments

where is the "two-wheel differential-drive-mobile robot" example?

tb12 gravatar image tb12  ( 2018-04-05 15:25:06 -0500 )edit

Question Tools

3 followers

Stats

Asked: 2018-03-14 04:01:25 -0500

Seen: 1,201 times

Last updated: Mar 14 '18